摘要： 针对机动车在检测和追踪过程中容易受到光照变化、目标遮挡以及天气变化等影响的问题,提出一种基于马尔可夫蒙特卡洛(MCMC)与多假设数据关联算法的多机动车检测和追踪算法。根据HOG特征模板匹配与MCMC运动状态估计,将外观模型和运动模型相结合,并保持全局数据关联。通过提高检测匹配阈值,降低运动估计误差,使算法满足精准性和实时性的要求。实验结果表明,该算法能够准确地估计机动车的运动状态,具有较高的检测准确率、精度与实时性,在正常光照条件下,检测和追踪的精度分别达到90%和85%以上。

关键词: 马尔可夫蒙特卡洛方法, 多假设数据关联, 多机动车检测, 多机动车追踪

Abstract: Aiming at the problem of illumination change,occlusion and weather condition in multiple vehicles detection and tracking process.This paper proposes the multiple detection and tracking algorithms,which are based on Markov Chain Monte Carlo(MCMC) method and Multiple Hypothesis Data Association(MHDA).The algorithms combine Histogram of Oriented Gradients(HOG) features template match with MCMC motion state estimation,and consist global data association.It improves the detection match threshold and decreases the motion estimation error to satisfy the accuracy and real-time.Experimental results show that algorithms have much higher accuracy and precision in detecting and tracking four kinds of vehicles and under normal light conditions,detection and tracking accuracy are 90% and 85% respectively.

Key words: Markov Chain Monte Carlo(MCMC)method, Multiple Hypothesis Data Association(MHDA), multiple motor vehicle detection, multiple motor vehicle tracking

中图分类号: 

• TP18